Project Proposal

Research Question

The aim of this analysis is to investigate diabetes prevalence over time and by country/region. The purpose is to identify countries and years with high diabetes prevalence.

Dataset Introduction

The dataset ‘DIABETES evolution of diabetes over time’ is a global dataset of diabetes prevelance from the years 1980 to 2014 and contains a total of 14,000 observations and 7 variables: - “Country/Region/World”, - e.g. “Turkey”, “Bangladesh”, “New Zealand” - “ISO”, - A region code for international standard denominations for country subdivisions - “Sex”, - Two factor variable “Men” or “Women” - “Year”, - Ranges from 1980 to 2014 - “Age-standardised diabetes prevalence”, - Calculated as a percentage - “Lower 95% uncertainty interval” and - “Upper 95% uncertainty interval”. Table 1 below shows the first six observations of the full dataset.

# Read in Data 

data_full <- read_csv("Data/Diabetes_data.csv")

## Rows: 14000 Columns: 7
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (3): Country/Region/World, ISO, Sex
## dbl (4): Year, Age-standardised diabetes prevalence, Lower 95% uncertainty i...
## 
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

data_full_head <- head(data_full)

kable(data_full_head, 
             caption = "Table 1: First Six Observations of the Full Diabetes Dataset",
             digits = 2)

Table 1: First Six Observations of the Full Diabetes Dataset

Country/Region/World	ISO	Sex	Year	Age-standardised diabetes prevalence	Lower 95% uncertainty interval	Upper 95% uncertainty interval
Afghanistan	AFG	Men	1980	0.04	0.02	0.09
Afghanistan	AFG	Men	1981	0.05	0.02	0.09
Afghanistan	AFG	Men	1982	0.05	0.02	0.09
Afghanistan	AFG	Men	1983	0.05	0.02	0.09
Afghanistan	AFG	Men	1984	0.05	0.02	0.09
Afghanistan	AFG	Men	1985	0.05	0.02	0.09

Dataset Description

The full dataset was reduced to 1000 observations through a random generation of row numbers. The variable “ISO” was removed as it was not necessary for analysis.** Figure 1** below shows the code used to tidy the full dataset into the reduced dataset.

include_graphics("Image/code_screenshot.png")

Figure 1: Code Screenshot of Data Tidying

Using the function str() the first 2 rows of the data is displayed to show the type of variables in the data set (numeric, character/factor etc.). The assessment requires a maximum of 5 variables, but both the “lower_95” and “upper_95” were kept as they work together.

head_data_2 <- head(data,2)
str(head_data_2)

## tibble [2 × 6] (S3: tbl_df/tbl/data.frame)
##  $ Country/Region/World: chr [1:2] "Russian Federation" "Czech Republic"
##  $ Sex                 : chr [1:2] "Women" "Men"
##  $ Year                : num [1:2] 2009 2013
##  $ diabetes_prevalence : num [1:2] 0.0779 0.0834
##  $ lower_95            : num [1:2] 0.0441 0.0448
##  $ upper_95            : num [1:2] 0.121 0.136

Data Summary

Two summary statistics were calculated for diabetes prevalence by “Year”. Table 2 shows the results of the summary statistics.

tail_data_summary <- tail(data_summary, 10)

kable(tail_data_summary, 
             caption = "Table 2: Mean and Standard Deviation of Diabetes Prevalence by Year (First 10 Rows)",
             digits = 3, 
      row_number(10))

Table 2: Mean and Standard Deviation of Diabetes Prevalence by Year (First 10 Rows)

Year	mean_diabetes	sd_diabetes	mean_upper95	sd_upper95	mean_lower95	sd_lower95
2005	0.108	0.057	0.147	0.072	0.045	0.045
2006	0.093	0.056	0.131	0.071	0.043	0.043
2007	0.083	0.048	0.116	0.061	0.036	0.036
2008	0.101	0.046	0.141	0.057	0.035	0.035
2009	0.096	0.058	0.136	0.075	0.043	0.043
2010	0.102	0.071	0.144	0.090	0.053	0.053
2011	0.089	0.048	0.133	0.064	0.033	0.033
2012	0.120	0.064	0.176	0.083	0.046	0.046
2013	0.094	0.031	0.150	0.047	0.019	0.019
2014	0.093	0.046	0.153	0.066	0.029	0.029

From Table 2 we can see an increasing trend in mean diabetes prevalence from 2005 to 2014. 2009 had the highest mean diabetes prevalence at 11.1% from the period 2005 to 2014, but also the highest standard deviation.

Visualisations

A figure was created using the ggplot2 R package and the option geom_point(). This is displayed in Figure 2:

Figure_2 <- ggplot(data = data_summary, aes(x = Year, y = mean_diabetes)) + 
  geom_point(alpha = 0.7) + 
  labs(title = "Figure 2: Mean Diabetes Prevalence Increases Over Time", 
       caption = "geom_smooth(` using method = 'loess' and formula = 'y ~ x'",
       subtitle = "Red Bars Represent Standard Deviation") + 
  xlab("Year") + 
  ylab("Mean Diabetes Prevalence") + 
  theme_minimal() + 
  geom_smooth() + 
  geom_errorbar(aes(ymin=mean_diabetes-sd_diabetes, ymax=mean_diabetes+sd_diabetes), colour="red", alpha=0.3)

ggplotly(Figure_2)

Figure 2 Summary

• There is a slight increase in mean diabetes prevalence from 1980 to 2014
• The standard deviation bars indicate a high dispersion of data

Conclusions

• When grouping diabetes prevalence by year, mean diabates prevalence appears to have a positive trend over time.
• This positive trend also includes fluctuations over years.
• 2009 has the highest diabetes prevalence, whereas 1981 has the lowest.
• The standard deviations for each year are high indicating a high dispersion of data.
• The next investigation would be to look at indivudal countries or grouping by sex.
• These results were also calculated from only 1000 entries of a full dataset containing 14,000 entries.
  • Possible error could be due to the random generation of rows and skew results.

Australian Diabates Trends By Sex

Australia_summary <- data_full %>%
  filter(`Country/Region/World` == "Australia")

Figure_3 <- ggplot(data = Australia_summary, aes(x = Year, y = `Age-standardised diabetes prevalence`, col = Sex)) + 
  geom_point(alpha = 0.8) + 
  labs(title = "Figure 3: Men have Higher Risk of Diabetes", 
       caption = "geom_smooth(` using method = 'loess' and formula = 'y ~ x'",
       subtitle = "Mean Diabetes Prevalence Has Increased Over Time") + 
  xlab("Year") + 
  ylab("Mean Diabetes Prevalence") + 
  theme_minimal() + 
  geom_smooth()

Figure_3

Figure 3 @ref(figure3) shows a trend of increasing mean diabates prevalence over time. Men have a noticeably higher mean than women. There is a steep increase from 1980 to 2000 and then a plateau. Data was only available up to 2010. It is unknown whether the plateua begins to trend downards.

Mean Diabetes Prevalence by Country

Australia_table_summary <- data_full %>%
  filter(`Country/Region/World` %in% c("Australia", "Germany", "China", "South Africa", "United States of America")) %>%
  select(-ISO) %>%
  group_by(`Country/Region/World`, Sex) %>%
  summarise(`Mean diabetes prevalence` = mean(`Age-standardised diabetes prevalence`))
  
kable(Australia_table_summary, 
             caption = "Table 1: First Six Observations of the Full Diabetes Dataset",
             digits = 3)

Table 1: First Six Observations of the Full Diabetes Dataset

Country/Region/World	Sex	Mean diabetes prevalence
Australia	Men	0.064
Australia	Women	0.047
China	Men	0.060
China	Women	0.061
Germany	Men	0.056
Germany	Women	0.040
South Africa	Men	0.069
South Africa	Women	0.097
United States of America	Men	0.065
United States of America	Women	0.054

Five random countries were selected to compare mean diabetes prevalence by year and sex. Table 3 @ref(table3) presents that in Australia, Germany and United States of America, men have a higher mean diabetes prevalence than women. Mean diabetes prevalence for men and women in China are very similar with men being 0.001 higher. Interestingly, women in South Africa have a higher mean diabetes prevalence than men.